基于cgroup的数据库内存超限冻结方法  

Freezing method for database based on cgroup when out of memory

作　　者：杨铸 高琳 YANG Zhu;GAO Lin(Beijing Eastern Jinxin Technology Co.,Ltd.Tianjin Hexi Branch,Tianjin 300210,China;Tianjin Beihai Communication Technology Co.,Ltd.,Tianjin 300384,China)

机构地区：[1]北京东方金信科技股份有限公司天津河西区分公司,天津300210 [2]天津市北海通信技术有限公司,天津300384

出　　处：《智能计算机与应用》2025年第1期75-80,共6页Intelligent Computer and Applications

摘　　要：分布式数据库作为一种大数据应用软件,需要使用有限的系统资源完成复杂并发计算和存储任务。海量的数据计算对内存需求也很大,为了多任务资源隔离,需要对不同任务组进行内存限制。当内存超限(OOM)时,操作系统默认强制杀死进程是高风险行为。为此,本文从一个简化应用开始,探求一种基于Linux内核控制群组(cgroup)功能的内存超限进程冻结与解冻技术,并应用于开源的Greenplum Database分布式数据库。As a big data application software,distributed database needs limited system resources to complete complex concurrent computing and storage tasks.Massive data computing also requires a lot of memory,and in order to isolate multi-task resources,it is necessary to limit memory for different task groups.When out of memory(OOM),the operating system defaults to forcing to kill processes as a high-risk behavior.Therefore,this research starts with a simplified application,explores a method of process freezing and unfreezing based on the Linux kernel control group(cgroup) feature when OOM,and then applies it to the open-source distributed database Greenplum Database.

关 键 词：大数据 分布式数据库 控制群组 内存 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]